74
Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing Liu

Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

  • Upload
    others

  • View
    5

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Aspect Extraction with Automated Prior Knowledge Learning

Zhiyuan (Brett) Chen Arjun Mukherjee

Bing Liu

Page 2: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Aspect Extraction

Extracting  aspect  terms

Page 3: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Aspect Terms

This  camera  takes  beautiful  pictures  but  its  price  is  higher  than  $200.

Page 4: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Aspect Terms

This  camera  takes  beautiful  pictures  but  its  price  is  higher  than  $200.

Page 5: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Aspect Extraction

Grouping  terms  into  categories

Extracting  aspect  terms

Page 6: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Grouping

PicturePhotoImageAspect  1 Aspect  2

PriceCostMoney

Page 7: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Aspect Extraction

Input:  A  review  collection

Output:  A  set  of  aspects(with  top  aspect  terms). Price

CheapCostMoneyPricy

BatteryLifeChargeAAAHour

Aspect  1 Aspect  2

Page 8: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic Models to Extract Aspects (e.g.,  Chen  et  al.,  2013;  Kim  et  al.,  2013;  Lazaridou  et  al.,  2013;  Mukherjee  and  Liu,  2012;  Moghaddam  and  Ester,  2011;  Sauper  et  al.,  2011;  Lin  and  He,  2009;  Titov  and  McDonald,  2008;  Lu  and  Zhai,  2008;)

Perform  both  extracting  and  grouping

A  topic  is  basically  an  aspect

Page 9: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Traditional Modeling Flow

M  DocsDomain  1

Page 10: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Traditional Modeling Flow

T  Topics

LDA

M  DocsDomain  1

Page 11: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Traditional Modeling Flow

T  Topics

LDA

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

Page 12: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Traditional Modeling Flow

T  Topics

LDA

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

Page 13: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Can we improve these topics by using them only?

Page 14: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Can we improve these topics by using them only? Fully automatic No other resources No human intervention

Page 15: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Our Proposed Algorithm

Page 16: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Topic  Base

Our Proposed Algorithm

Page 17: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Knowledge  BaseLearn  Knowledge  Automatically

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Topic  Base

Our Proposed Algorithm

Page 18: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Knowledge  BaseLearn  Knowledge  Automatically

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Topic  Base

Our Proposed Algorithm

a)  Existing  Domains

Page 19: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL (Automated !Knowledge LDA)!

Knowledge  BaseLearn  Knowledge  Automatically

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Topic  Base

M  DocsDomain  1

T  Topics

AKL

M  DocsDomain  2

T  Topics

AKL

M  DocsDomain  N

T  Topics

AKL

Our Proposed Algorithm

a)  Existing  Domains

Page 20: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Knowledge  BaseLearn  Knowledge  Automatically

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Topic  Base

Our Proposed Algorithm

b)  New  Domain

Page 21: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Knowledge  BaseLearn  Knowledge  Automatically

M  DocsDomain  1

T  Topics

LDA

M  DocsDomain  2

T  Topics

LDA

M  DocsDomain  N

T  Topics

LDA

Topic  Base

M  DocsDomain  N+1

T  Topics

AKL

Our Proposed Algorithm

b)  New  Domain

Page 22: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Why don’t we merge documents from different domains and run LDA?

Page 23: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Run LDA on Merged Data

Number  of  Topics

Topic  belongs  to  which  domain

Scalability

Page 24: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

M  Docs

M  Docs M  Docs

M  Docs

M  Docs

Run LDA on Merged Data

Page 25: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Run LDA on Merged Data

Run  LDA

Page 26: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Our Proposed Algorithm Run  LDA Run  LDA Run  LDA

Run  LDA

Run  LDA

Page 27: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

T  Topics

T  Topics T  Topics

T  Topics

T  Topics

Our Proposed Algorithm

Page 28: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Our Proposed Algorithm

Learn  Knowledge

Page 29: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Knowledge

Our Proposed Algorithm

Knowledge Knowledge

Knowledge

Knowledge

Page 30: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Our Proposed Algorithm Run  AKL Run  AKL Run  AKL

Run  AKL

Run  AKL

Page 31: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Multiple  Senses

KnowledgeReliability

Learn Knowledge Automatically

Page 32: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Multiple  Senses

KnowledgeReliability

Learn Knowledge Automatically

Page 33: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

{Light,  Bright}{Light,  Luminance}

{Light,  Weight}{Light,  Heavy}

Light

Multiple Senses

Page 34: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Existing  Models  with  Multiple Senses

Assume  single  senseDF-‐‑‒LDA  (Andrzejewski  et  al.,  2009)

User  specified  multiple  sensesMC-‐‑‒LDA  (Chen  et  al.,  2013)

Automatically  distinguish  senses  when  extracting  knowledge

Page 35: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Multiple  Senses

KnowledgeReliability

Topic  Clustering

Learn knowledge Automatically

Page 36: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic Clustering

A  topic  represents  words  with  similar  meaning  (but  noisy)

Group  topics  with  similar  sense  into  one  cluster

Different  senses  of  a  word  should  be  split  into  different  clusters

Page 37: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Multiple  Senses

KnowledgeReliability

Topic  Clustering

Learn knowledge Automatically

Page 38: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic Overlapping

Every  product  domain  has  price.

Most  electronic  domains  have  battery.

Some  electronic  domains  share  screen.

Page 39: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Example BatteryLifePictureCharge

BatteryPriceLifeSize

BatteryChargeAAAScreen

D1                                        D2                                    D3

Page 40: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Example

BatteryLifePictureCharge

BatteryPriceLifeSize

BatteryChargeAAAScreen

D1                                        D2                                    D3

Two  words  together  at  least  2  times

Page 41: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Example

BatteryLifePictureCharge

BatteryPriceLifeSize

BatteryChargeAAAScreen

D1                                        D2                                    D3

Two  words  together  at  least  2  times

{Battery,  Life}  and  {Battery,  Charge}

Page 42: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Multiple  Senses

KnowledgeReliability

Topic  Clustering

Frequent  Itemset  Mining

Learn knowledge Automatically

Page 43: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Frequent Itemset Mining (FIM)

Each  topic  is  a  transaction

Find  frequent  patterns  satisfy  minimum  support  thresholds

Each  pattern  contains  2  terms

Page 44: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Knowledge Representation

In  the  form  of  knowledge  clusters  (KC)

Each  KC  has  a  list  of  frequent  2-‐‑‒patterns

KC1:  {battery,  life},  {battery,  charge},  {battery,  hour},  {charge,  hour}

Page 45: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL (Automated Knowledge LDA)

Incorporate  Knowledge

Wrong  Know.  Towards  Domain

Page 46: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Model

Add  variable  cIncorporate  Knowledge

Wrong  Know.  Towards  Domain

Page 47: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Plate Notation

c:  knowledge  cluster

Page 48: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Plate Notation

c:  knowledge  cluster

Page 49: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Plate Notation

c:  knowledge  cluster

Page 50: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Plate Notation

c:  knowledge  cluster

Page 51: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Model

Add  variable  c

GPU  Model

Incorporate  Knowledge

Wrong  Know.  Towards  Domain

Page 52: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic  0

price

LDA with SPU (Simple Pólya Urn Model)

Page 53: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic  0

price price

LDA with SPU (Simple Pólya Urn Model)

Page 54: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic  0

price

AKL with GPU (Generalized Pólya Urn Model)

Page 55: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Topic  0

price price

cheap

{price,  cheap}AKL with GPU (Generalized Pólya Urn Model)

Page 56: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Model

Add  variable  c

GPU  Model

Incorporate  Knowledge

Wrong  Know.  Towards  Domain

Page 57: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Wrong Know. Towards Domain Wrong  because  of  TM  mistakes{Price,  Picture}

Wrong  towards  a  particular  domain  {Light,  Bright}{Light,  Weight}

Page 58: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Model

Add  variable  c

GPU  Model

Co-‐‑‒Document  Frequency  Ratio

Incorporate  Knowledge

Wrong  Know.  Towards  Domain

Page 59: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Co-Document Frequency Ratio

Page 60: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Co-Document Frequency Ratio

Estimated  in  the  current  domain

Page 61: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Co-Document Frequency Ratio

Estimated  in  the  current  domain

{Price,  Cheap}{Price,  Image}

Page 62: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Evaluation

Page 63: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Evaluation 36  product  domains.  Each  domain:1000  Reviews15  Topics

EvaluationHuman

Objective

Page 64: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Model Comparison

LDA  (Blei  et  al.,  2003)

GK-‐‑‒LDA  (Chen  et  al.,  2013)

MC-‐‑‒LDA  (Chen  et  al.,  2013)

Page 65: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Model Comparison

LDA  (Blei  et  al.,  2003)

GK-‐‑‒LDA  (Chen  et  al.,  2013)

Feed  them  with  the  knowledge  from  our  algorithm

MC-‐‑‒LDA  (Chen  et  al.,  2013)

Page 66: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Objective Evaluation

-1510

-1490

-1470

-1450

-1430

0 1 2 3 4 5 6

Topi

c C

oher

ence

AKL GK-LDAMC-LDA LDA

Page 67: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Example Aspects

Page 68: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Human Evaluation

0.6

0.7

0.8

0.9

1.0

Camera Computer Headphone GPS

Precision

@ 5

AKL GK-LDA MC-LDA LDA

Page 69: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Human Evaluation

0.6

0.7

0.8

0.9

1.0

Camera Computer Headphone GPS

Precision

@ 1

0

AKL GK-LDA MC-LDA LDA

Page 70: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Number of Topic Clusters

-1510-1490-1470-1450-1430

20 30 40 50 60 70Top

ic C

oher

ence

#Clusters

Page 71: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Conclusions

To  extract  better  aspects

Learn  knowledge  automatically

AKL:  Leverage  automated  knowledge

Page 72: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Multiple  Senses

KnowledgeReliability

Learn knowledge Automatically

Topic  Clustering

Frequent  Itemset  Mining

Page 73: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

AKL Model

Add  variable  c

GPU  Model

Co-‐‑‒Document  Frequency  Ratio

Incorporate  Knowledge

Wrong  Know.  Towards  Domain

Page 74: Aspect Extraction with Automated Prior Knowledge Learningzchen/presentations/ACL...Aspect Extraction with Automated Prior Knowledge Learning Zhiyuan (Brett) Chen Arjun Mukherjee Bing

Q&A